There is a concern that LLM assisted learning in schools and universities is used as a crutch to offload the more difficult aspects of learning. In this project we investigate if this impoverishment really occurs and explore how insights from pedagogical psychology can be leveraged to mitigate these risks and optimize human-AI collaboration.
Large language models (LLMs) may have great potential to facilitate and support the learning process. However, there is a growing concern that LLMs may lead to offloading the core aspects of learning in some students and therefore rather hinder than assist in achieving educational goals. This project is aimed at answering the following three questions:
- How often does this offloading of difficult aspects occur during learning, and how does the use of these LLMs evolve during the first year at university?
- Can AI technology be designed to enhance human learning efficiency?
- Can established findings from pedagogical psychology, such as elaboration, be applied to facilitate LLM utilization in a way that benefits learning?
To address these questions, we will conduct three studies. First, a one-year exploratory longitudinal observational study will examine how students AI utilization interacts with academic outcomes, motivation, and learning strategies over time. Second, we conduct an experimental laboratory study with a controlled group comparison to investigate the impact of different AI technologies (LLM and Retrieval-Augmented Generation) on academic performance and brain activity using electroencephalography (EEG). Finally, we will investigate how training in elaborative learning strategies, the type of learning partner (human or LLM), and testing delays interact to affect a student's retention score and brain activity.
All in all, these studies provide insights into how AI and learning strategies can be integrated into higher education.
